---
title: Data collections
slug: q64g0205
canonical_url: https://docs.coveo.com/en/q64g0205/
collection: data-collections
source_format: adoc
---
# Data collections

Data collections are intended for administrators and developers who want to store and manage structured business data in Coveo.
This article provides an overview of [data collections](https://docs.coveo.com/en/q65b0594.md), including their purpose, how they relate to other Coveo [resources](https://docs.coveo.com/en/2820.md), and the types of data they can contain.
It also introduces the key concepts and workflows involved in working with [data collections](https://docs.coveo.com/en/q65b0594.md).

> **Note**
>
> Data collections are actively evolving.
> Their capabilities and supported use cases may expand over time.

## What's a data collection?

A [data collection](https://docs.coveo.com/en/q65b0594.md) is a container used to store and manage structured business data provided by customers.
Unlike sources, [data collection](https://docs.coveo.com/en/q65b0594.md) [items](https://docs.coveo.com/en/210.md) aren't designed to be searched or used directly in search or recommendation features.
Instead, they store datasets that provide context for Coveo features, such as [Coveo Machine Learning (Coveo ML)](https://docs.coveo.com/en/188.md) [models](https://docs.coveo.com/en/1012.md).

For example, suppose you have a Coveo commerce implementation that automatically logs [events](https://docs.coveo.com/en/260.md) [using Event Protocol](https://docs.coveo.com/en/3188.md#protocols).
If you also collect in-store purchase data that provides additional insight into customer purchase patterns, you might want your [Coveo Machine Learning (Coveo ML)](https://docs.coveo.com/en/188.md) models to use that information too.
You can use an offline purchase data collection to submit this information in a format that aligns with the data collected online.

Because the features they support require consistent, high-quality data, Coveo predefines and enforces a data structure for each data collection type.
Data collections offer flexibility in the data you can submit while providing a single mechanism for managing that data.

At this time, you must use the private Data Collections API to [create a data collection](https://docs.coveo.com/en/q65f3128.md).
You can then use Stream API requests to [add, update, and delete items](https://docs.coveo.com/en/q6980563.md) in your data collection.

![Data collections in the Coveo Platform architecture | Coveo Platform](https://docs.coveo.com/en/assets/images/data-collections/data-collections-architecture.png)

| 1 | Before data ingestion, create a data collection using the private Data Collections API. |
|---|---|
| 2 | Use Stream API requests to add, update, and delete items in your data collection. |
| 3 | Added and updated items are validated against the [predefined schema](https://docs.coveo.com/en/q6980563.md#validation-schemas) associated with the data collection's `dataCollectionType`. Use the [**Log Browser**](https://platform.cloud.coveo.com/admin/#/orgid/content/log-browser/) ([platform-ca](https://platform-ca.cloud.coveo.com/admin/#/orgid/content/log-browser/) | [platform-eu](https://platform-eu.cloud.coveo.com/admin/#/orgid/content/log-browser/) | [platform-au](https://platform-au.cloud.coveo.com/admin/#/orgid/content/log-browser/)) to [review validation errors](https://docs.coveo.com/en/q6980563.md#review-ingestion-logs) and other issues with ingested data. |
| 4 | Data collection items are stored for usage by internal services. |
| 5 | (Coming soon) [Coveo Machine Learning (Coveo ML)](https://docs.coveo.com/en/188.md) [models](https://docs.coveo.com/en/1012.md) will be able to use data collection items as training data, allowing you to provide additional context and signal to Coveo Machine Learning features. |

## When to use data collections

As an initial use case, [data collections](https://docs.coveo.com/en/q65b0594.md) can be used to send offline purchase data to the [Coveo Platform](https://docs.coveo.com/en/186.md) for commerce customers.
This will soon allow recommendation implementers to selectively train [Coveo Machine Learning (Coveo ML)](https://docs.coveo.com/en/188.md) [Product Recommendation (PR)](https://docs.coveo.com/en/3132.md) [models](https://docs.coveo.com/en/1012.md) on offline purchase data and configure a [recommendation strategy](https://docs.coveo.com/en/p85e0425.md) that more accurately reflects purchasing behavior across online and offline channels.

> **Note**
>
> Among further use cases, Coveo plans to use [data collections](https://docs.coveo.com/en/q65b0594.md) to manage user profiles for personalized search and recommendations.

## What's next

[Create a data collection](https://docs.coveo.com/en/q65f3128.md).